Multi-scale real-time prediction of indoor thermal environmental in Sichuan Plateau ethnic minorities traditional dwelling based on Wavelet-enhanced Graph Recurrent Network

Accurate prediction of indoor thermal environment is crucial for achieving good indoor environmental quality and efficient energy management in traditional dwellings. In this study, a multi-scale real-time prediction method for indoor thermal environmental of traditional houses of ethnic minorities in the Sichuan Plateau based on Wavelet-enhanced Graph Recurrent Network (WAGRNet) was proposed. To reduce the impact of external factors on test accuracy, we selected the same weather conditions for the 2024 transition season from May 11 to 17 and summer from July 19 to 25 conduct this test. The measurements are centered on three parameters: the indoor air temperature, relative humidity, wind speed which were monitored continuously for 24 hours a day across one representative traditional dwelling. WAGRNet was validated by comparing its prediction performance against baseline models using MSE, RMSE, and MAE metrics. In the summer prediction task, WAGRNet achieved MSE, RMSE, and MAE of 2.307 ± 0.373, 1.515 ± 0.121, and 1.068 ± 0.097, respectively, yielding relative error reductions of approximately10.5% (MSE) and 7.4% (MAE) compared to the best baseline model. In the transition season prediction task, WAGRNet maintained its superior performance with MSE, RMSE, and MAE of 7.642 ± 2.560, 2.732 ± 0.474, and 1.873 ± 0.284, respectively, outperforming the best baseline (8.402 ± 3.051, 2.862 ± 0.510, 1.991 ± 0.320) with relative error decreases of about 9.0% (MSE) and 5.9% (MAE). Additional ablation experiments, five-seed reliability analysis, confidence intervals, statistical tests, and inference-time benchmarking further confirmed the contribution of the proposed modules and the computational feasibility of WAGRNet for real-time prediction indoor thermal environment. This study presents a field-measured, data-driven framework for predicting the indoor thermal environment of traditional dwellings on the Sichuan Plateau, thereby facilitating their modernization and preservation, and providing a scientific basis for improving the quality of residents' living conditions and sustainable development.